Performance Evaluation and Efficiency Analysis of …banker/dea2009/paper/Behera.pdfIndian economy, which till recently ... Performance evaluation and efficiency analysis of the thermal

Performance Evaluation and Efficiency Analysis of the Coal Fired

Thermal Power Plants in India

Santosh K Behera1*

, Jamal A Farooquie2, Ambika P Dash

3

1Sr. Manager (Corporate Planning) , NTPC Ltd., India

2Associate Professor, Department of Business Administration, Aligarh Muslim University, India.

3 Professor, (Finance & Strategy), Power Management Institute, NTPC Ltd., India

Abstract

Power is the key input for most of the industrial, agricultural and social establishments.

Indian economy, which till recently grew at a faster rate of above 9 per cent, faces power

shortage. Several structural as well as institutional reforms have been undertaken by the

Government of India to mitigate the perennial problem of power shortage. Despite the

regular additions of power generating capacity, the gap between the generation of power

and its demand has always been widening. This paper attempts to investigate whether this

gap can be reduced through making the existing power plants more efficient. Efficient

power generation is expected to make more power available at a lower cost for economic

and other activities, which in turn shall make the country more competitive. The focus of

the study is on the coal fired thermal power plants in the country. Thermal power in India

accounts for about 64.6 per cent of the total power generation capacity. Out of which, the

contribution of coal-fired plants has been 53.3 per cent. Data envelopment analysis (DEA)

has been used to estimate the relative performance of the coal fired power-generating plants

in India and explore the key determinants of the inefficient units.

Keywords: Data envelopment analysis (DEA), Performance evaluation, Power Generation,

Coal fired thermal power plants.

*

* Corresponding Author, email: [email protected]. The opinion expressed in the paper is purely the

personal opinion of the authors and not of the organizations they represent.

Performance evaluation and efficiency analysis of the thermal power plants in India

2

1. Introduction

India has an installed capacity to generate 1,47,965.41 MW of electricity along with the

captive generation capacity of 19509.49 MW connected to the grid. Out of this, the thermal

power contributes 93,725.24 MW (64.6%). Based on the type of primary fuel used, thermal

power plants are of three types viz. coal based, gas based and diesel based. Currently the

installed capacity of coal based, gas based and diesel fired thermal power plants is

77648.88MW, 14876.61MW and 1199.75MW, respectively (as on 31-03-2009, Source:

Ministry of Power, Govt. of India Website, http://powermin.nic.in – last visited on 12-05-

2009). The trend in the growth and composition of installed capacity is depicted in Figure-

1. The coal based thermal power generating units dominate Indian power sector

contributing to 53.3% of installed capacity and are managed in three sectors – state sector,

comprising of the State Electricity Boards (SEBs) and their unbundled generation units;

central sector, comprising of NTPC, Damodar Valley Corporation (DVC) and Neyveli

Lignite; and private sector, comprising of Tata Power, Reliance Energy, CSES etc.

0 20000 40000 60000 80000

100000 120000 140000 160000 Ca

pacity (in MW)

6th

Plan 7th

Plan Ann.

Plan-2 8th

Plan 9th

Plan 10th

Plan 11th

Plan

Period

Growth of installed capacity

(as on 31.03.2009)

Coal Gas Diesel Nuclear Hydro Renewables

[Figure -1: Growth of Installed Capacity during the period 01-04-1980 to 31.03.2009]


3

The thermal power generation capacity is built over years and consists of units of different

capacities ranging from 20MW to 500MW. The 660MW capacity units are being

introduced in the central sector for the first time by NTPC. NTPC is the biggest operator in

the country having an installed capacity of 30144MW from its 77 coal fired and 32 gas

fired power generating units. The recent initiative of Government of India to enhance

thermal capacity in the form of Ultra Mega Powers Projects (UMPP) are expected to have

unit sizes of 800MW (http://powermin.nic.in – Last visited on 12-05-2009). Indian power

generation sector is the fifth largest in the world and has generated 723.469 Billion Units

with average cost of supply @Rs2.76 / unit during the year 2008-09.

The thermal power generating units primarily consume scarce and non-renewable fossil

fuels. Excess consumption by one plant has the cascading effect of increasing the

production cost of that unit as well as depriving other remaining plants of this limited

natural resource. Indian power sector faces acute shortage of coal and as a result several

units are shutdown. During 2005-06, power stations lost generations of 1653.5 MUs due to

shortage of coal which further aggravated during 2008-09 resulting loss of generation of 9

Billion Units (up to Dec’2008). Reduction in the consumption of coal/ oil by one unit shall

make available is to other units thereby reducing cost of electricity which being an input to

most industrial and social process shall provide more competitiveness to their products /

services. Power stations consume a portion of electricity generated to power the auxiliary

equipments called Auxiliary Power (APC). For the year 2005-06 this varied from 5.59% to

16.23% with average of 8.44%. Any reduction in APC shall make more power available to

the grid for fuelling the national economy. In this context, estimation of the productive

efficiency of the power generating units is important as this gives an idea of the current

level of productivity of the power generation process, the consumption levels of various

inputs and directions as well as quantum of possible improvement possible. In this paper,

attempt was made to quantitatively estimate the relative performance level of the power

generation units based on multiple inputs and analyse the determinants of less efficient

units. The rest of the paper is organized as follows: section 2 reviews the literature on the

productivity studies in the power sector, multi criteria quantitative techniques used for the

productivity measurement, Data Envelopment Analysis (DEA) and its applications, section

3 reviews the performance parameters of the plants under study and current approaches to


4

performance evaluation followed by section 4 which explains the concepts of DEA and

section 5 which focuses on data analysis and discussion. The last section, Chapter 6 details

conclusions and the limitations of the current study and scope for further research.

2 Review of Literature:

‘When you can measure what you are speaking about and express it in numbers, you know

something about it’ says Kelvin. Sumanth (1998) lists Measurement, Evaluation, Planning

and Improvement as the four phases of a productivity process. Productivity measurement

thus holds the key to any productivity management exercise. Increased productivity

enhances the competitive advantages of a firm in the form of decreased product cost,

improved product quality leading to improved market share and profit. Key to the

productivity enhancement lies in identifying areas of potential productivity improvement.

Several techniques have been used to quantitatively estimate the productivity levels of

various processes. The classic measure of productivity as the ratio of output to input, which

does very well for the single input and output processes, fares badly with the increasing

complexity of the modern day business, processes which in reality consumes multiple

inputs to produce a variety of outputs. A production function is defined as a relationship

between the maximal technically feasible output and the inputs needed to produce that

output – Shephard (1970). Mishra (http://ssrn.com/abstract=1020577 – last visited on 20-

05-2009) traces the history of production functions, which has been formulated over the

years to unravel the underlying relationship between the inputs and outputs. Technical

efficiency (TE) of a firm reflects its ability to minimize usage of inputs to produce a given

amount of output. The firm, which uses the least input, is called technically efficient and

has a TE score of 100%. Over the years Stochastic Frontier Analysis (SFA) and Data

Envelopment Analysis (DEA) has emerged as the preferred quantitative techniques for

performance measurement based on multiple input and output criteria. SFA is the result of

three independent models proposed around same time in 1977 by Meeusen and van den

Broeck (MB), Aigner, Lovell and Schimdt (ALS) and Battese and Corra (BC). SFA is a

parametric method and requires assumption of a functional form of the productivity process

and capable of handling stochastic error. The story of DEA dates back to Rhodes’s PhD

thesis which led to the publication of the CCR model in Cooper (1978). DEA is a linear


5

programming based non-parametric method used to measure the relative performance level

of homogenous firms called Decision Making Units (DMU) like banks, hospitals,

municipalities etc. The initial (CCR) model without taking scales of operation into account

was subsequently modified by incorporating variable return to scale in the BCC model by

Banker (1984).

The basic DEA models has been augmented by a) Additive Model – which combines input

and output orientations, b) Slack Based Measure – making the additive model unit

invariant, c) Hybrid Model – unifying radial and non radial measures, d) Free Disposal

Hull (FDH) model, e) Super Efficiency Model – ranking of efficient units, f) Models with

Restricted Multipliers – to incorporate subjective assessments by way of weight

restrictions, g) Non-controllable, Non-discretionary and Bounded Variable models.

DEA has been used with other tools like Regression, Principal Component Analysis (PCA)

– Adler (2001), Adler(2009), Stochastic Frontier Analysis (SFA) – Li (1998), Huan and Li

(2002), Kuosmanen (2006), Fuzzy logic – Kao (2000) and Guo (2001), Artificial Neural

Networks (ANN) - Wu (2004), Celebi (2008) and Emrouznejad (2009) over the years to

take care of the diversities and complexities in the real world problems. The application of

DEA from performance measurement has been expanded to other areas like policy studies -

Gurgen(2006) , Iimi(2003), Delmas (2003) , Toba(2003) and Arocena (1999),

Benchmarking , Checking a virtual merger, Comparison of business models, Site

Selections etc. - Cooper (2007). Emrouznejad (2001) provides an extensive collection of

DEA literature.

Several studies have been conducted world wide to estimate the performance level of

electric power industry. Golany et al. (1994) studied the relative performance level of

thermal power plants in Israel. Chitkara (1999) and Arocena (1999) attempted to measure

the efficiency of power generation units in India and Spain respectively. Diewert and

Nakamura (1999) examined 77 plants in 28 countries funded by World Bank for

benchmarking and the measurement of best practice efficiency. Olatubi (2000), Lam (2001)

and Sueyoshi (2001) studied the performance levels of electric utilities in US, China and

Japan respectively.


6

Shanmugam (2005) analyzed the efficiency of 56 coal based thermal power generation

stations in India during the period 1994-95 to 2001-02 using Stochastic Frontier Analysis

(SFA). They have used the capital employed, specific coal consumption, specific secondary

oil consumption, auxiliary power consumption and power generated as variables and found

that the technical efficiency varies from 46% to 96% and is time invariant. They have also

found that 22 out of 56 power stations analyzed have TE below 70%. Nag (2006)

developed a framework to estimate the carbon base line for power generation in India till

the end of 11th

five-year plan period (2010-11) based on the Specific Coal Consumption

(SCC) and APC. Dash, Behera and Rath (2008), while exploring alternative matrices for

India’s future power demand have observed that there is substantial scope for improvement

of the performance of the thermal power plants and suggested for aggressive action plans

for augmenting the current output levels.

3 Performance Parameters and Measurement Systems:

Coal based thermal power plants are capital intensive and takes almost 6 years from

concept to commissioning. At the current level, the tentative cost of these units varies

between Rs 40 to Rs 50 Million / MW (http://cea.nic.in – last visited on 20-05-2009). The

key input in any performance measurement exercise of these plants should be the cost of

capital. Since the capacities are built over years and cost of capital is not available for the

many of the operational plants, installed capacity is taken as an indicator of capital.

Shanmugam (2005) considered capacity as capital input. These plants primarily use coal

and oil as fuels and electricity to power auxiliary equipments. The amount of coal and oil

consumed to generate one unit (Kilo Watt Hour -KWH) of electricity are called Specific

Coal Consumption (SCC) and Specific Fuel Oil Consumption (SFOC). APC is considered

as a deemed input and is a part of the output. Other important inputs to the plants are

maintenance expenditure, which includes employee costs, inventory costs and other costs.

Since most of the power generating units are in the state sector, the employee costs,

inventory costs, profit earned are either not computed separately or not available in the

public domain.

The Operational Availability Factor (OAF) of a plant is the percentage of time a plant is

available for generation during a period. OAF is arrived at by excluding the duration for

which the plant is not available because of various outages. Cook (2005) categorized


7

outages into 4 categories which are long duration planned maintenance – usually for major

overhauls, short duration maintenance outage - for minor routine maintenance, unscheduled

forced outage – due to equipment failure with some prior warning and sudden outages -

forced outage without prior warning. In India the performance review by Central

Electricity Authority (CEA - http://cea.nic.in ) captures outages in two categories. While

the Planned Maintenance (PM) includes planned outage and maintenance outage, the

Forced Outage (FO) includes forced outage and sudden outage. FO ignites public opinion,

interrupts business operations, and generally reflects negatively on the organization and

should play a direct role in any measure of efficiency Cook (2005). In absence of explicit

maintenance cost information, PM and FO figures can be considered as an indicator of the

maintenance and opportunity costs. Since it is possible to reduce the APC, PM and FO

figures by means appropriate managerial intervention in the form of improved operation

and maintenance practices, and going by the argument of lower is the better for inputs,

APC, PM and FO can be considered as deemed inputs like Capacity, SCC and SFOC.

The important output parameters of thermal power plants are the amount of electricity

generated by a plant during a period in India this is usually measured in terms of Plant

Load Factor (PLF) which is the ratio of actual generation to theoretical possible generation

during a period which may be a day, month, quarter or a year.

Descriptive statistics of SCC, SFOC, APC, PM, FO and PLF is detailed in table-1.

a. Currently practice of performance evaluation of the thermal power units, is

based on ratio analysis. In this method, ratio of an output to an input (PLF, OAF) or an

input to an output (SCC, SFOC and APC) is computed and used as a performance indicator

(PI). Various PIs like PLF, OAF, SCC, SFOC, APC, PM, FO etc are computed and

indicated separately. While PIs provides useful information on the performance of a unit on

individual pairs of inputs and outputs, they are problematic when used to gain an overall

view of the unit’s performance - Thanassoulis (1996). While ratios are easy to compute,

which in part explains their wide appeal, their interpretation is problematic, especially

when two or more ratios provide conflicting signals. Therefore ratio analysis is often

criticized on the grounds of subjectivity, because an analyst must pick and choose ratios in

order to assess the overall performance of a firm Malhotra (2008). To foster the

competitive spirit amongst various power stations so as to encourage them to improve the


8

operational performance, Govt. of India has introduced several award schemes for this

sector formulated by CEA. Till 2003-04 separate award schemes were there for important

operational parameters viz PLF, SFOC, and APC. From the year 2004-05, important

operational parameters like specific fuel oil consumption, APC, peak PLF and SHR were

included to calculate a composite score. In October 2008, CEA (2008) proposed to include

design SHR in place of normative SHR in the performance calculation. The weight matrix

decided are [0.50 0.15 0.15 0.20] for the performance parameter matrix [Peaking_PLF

Station_Heat_Rate Specific_Fuel_Oil_Consumption Aux._Power_Consumption]. While

this methodology of computing a unified performance index is quite simple and relatively

easy, it raises a host of questions. Are the so-called efficient units truly efficient because of

their performance parameters or purely because of favorable weight matrix? How much of

the efficiency ratings are due to the weights and how much inefficiency is associated with

the observations? Cooper [2007]. Should the weight matrix be decided a priori or it should

be derived from the performance matrix.

4 Data Envelopment Analysis (DEA):

The primal version of DEA is called the multiplier version and involves discovering the

optimal set of weights for the inputs and outputs that maximizes the efficiency of the DMU

relative to other DMUs. The efficiency of a DMU which is defined as the ratio of the

weighted sum of outputs to weighted sum of inputs is maximized such that, the efficiency

of all other DMUs lie between 0 and 1. Maximizing the ratio involves fractional

programming, which can either be achieved by maximizing the numerator or minimizing

the denominator by setting the other to 1. For a set of N DMUs, consuming I inputs to

produce J outputs, the multiplier version involving maximization of the weighted sum of

outputs is represented as:

∑=

=

J

1j

jmv zmax jmy

Subject to

1 uI

1i

im =∑=

imx


9

- vJ

1j

jm∑=

jny 0 uI

1i

im ≤∑=

inx ; n = 1,2,K,N

vjm, uin ≥ ε ; i=1,2,K,I ; j= 1,2,K,J

Where

xim and yjm are the ith input is the jth output of the mth DMU

uim and vjm is the weight associated with xim and yjm

xin and yjn are the ith input and jth output of the nth DMU

The dual of the primal version is called the envelopment version and involves creating a

hypothetical DMU from the linear combination of the existing real DMUs that either

consumes less inputs to produce at least the same output (input oriented) or produces more

output without requiring additional inputs (output oriented).

If it not feasible to create a hypothetical DMU, then the DMU under evaluation is termed

efficient and the loci of such efficient units define the efficiency frontier. Else the DMU

under study is inefficient and the targets for the hypothetical DMU can be set for real

DMU. The quantity of input contraction or output augmentation, which can be achieved to

pull (input oriented) or push (output oriented) the inefficient units to the frontier, indicates

the degree of inefficiency and the margin for improvement. The input oriented model aims

to contract the input levels to produce at least the current output levels and is given by:

Min θm

θ,λ

Subject to

Yλ ≥Ym ; Xλ ≤ θm Xm; λ ≥ 0; θm free

The output oriented version is:

Max φm

Φ,µ

Subject to


10

Yµ ≥φmYm ; Xµ ≤ Xm ; µ ≥ 0; φm free

The strength of DEA lies in discovering the optimal set of weights for each DMU

from the performance of the DMUs itself, eliminating the subjective bias involved in

selecting them. DEA is capable of identifying the sources and amounts of inefficiency and

thus possible improvements, set rational targets and pin point bench mark members which

can be the sources of best practices for subsequent performance improvement initiatives.

All this is possible based on the actual achievements of individual units and not on the

theoretical estimates, without requiring assumption of a functional form for the production

function.

5 Data Analysis and Discussion:

The performance data of various power generating units is being compiled by CEA in its

Annual Thermal Performance Reviews. In this study the performance data for the year

2005-06, of 74 generating stations having an installed capacity of 62309MW out of the

commissioned capacity of 67284MW (as on 31-03-2006) (92.60 %) is taken into account –

CEA (2006). The review presents some of the unit level performance data like PLF, OAF,

PM, FO etc and few other plant level data APC, SCC, SOC. Even though unit level

analysis would have provided more focused results, to capture more parameters like APC,

SCC and SOC in the study, we considered individual plants as DMUs. Performance data

for some of the stations are not available in the report and attempt was made to collect the

data from individual generating stations. The performance data for some of the parameters,

which could not be collected from power stations, is interpolated from the historical

published data. The descriptive statistics of important operational performance parameters

is detailed in table-1.

Descriptive Statistics

N Range Minimum Maximum Mean

Std.

Deviation

APC 74 10.64 5.59 16.23 9.6572 2.28806

Capacity 74 2970.00 30.00 3000.00

842.013

5 635.80318

FO 74 51.53 .00 51.53 10.5726 11.56606

OAF 74 94.23 4.71 98.94 79.2131 19.57541

PLF 74 96.80 2.82 99.62 68.4249 24.03707

PM 74 95.29 .00 95.29 10.2143 14.06498

SFOC 74 38.85 .10 38.95 3.1165 5.60780


11

N Range Minimum Maximum Mean

Std.

Deviation

SCC 74 .63 .46 1.08 .7415 .13459

Valid N

(listwise) 74

Table – 1: Descriptive Statistics of Performance Data for the year 2005-06

Considering the large variation in installed capacities of power plants ranging from 30 MW

to 3000 MW, BCC- model to account for Variable Return to Scale is used so that the plants

are compared among the comparables. To use BCC model, the ratio data (in which scale

information is lost) is converted back to absolute values of the input and output parameters.

The analysis was done using six inputs – Capacity in MW, Coal Consumed in MT, Oil

Consumed in KL, APC in MU, PM and FO in deemed MUs and power generated in MUs.

DEA analysis for the 74 generating stations the performance parameter was performed

using DEAP 2.1 with input oriented envelopment model under variable return to scale.

With 6 inputs and 1 output, as a rule of thumb the DMUs should be more than max {6 X 1,

3(6+1)} i.e. 18 - Cooper (2007, page 284). The number of DMUs studied is 74, which are

quite good. As such with 74 DMUs under study, the number of input and output parameters

can be extended far beyond. The results are detailed below:

1. The CRS TE, VRS TE, scale efficiency, Peer Count, Return to scale (RTS) along

with input targets for the 74 generating stations is listed in table-2. Henceforth VRS TE is

indicated as efficiency unless otherwise specified.

DMU No

DMU Capacity CRS TE

VRS TE

RTS Scale Efficiency

Peers

1 Ahemadabad 420 1 1 - 1 1

2 Amarkantak 60 0.53 1 IRS 0.53 2

3 Amarkantak

Ext

240 0.563 0.631 IRS 0.891 68 , 41 ,52, 61

4 Anpara 1630 0.829 0.864 DRS 0.959 14, 73

5 Badarpur 720 0.866 0.873 DRS 0.992 73, 14

6 Bandel 540 0.827 0.829 IRS 0.998 1, 41, 19, 68

7 Barauni 320 0.448 0.479 IRS 0.935 68, 1, 41

8 Bakeshwar 630 0.957 0.957 - 1 53, 14, 19, 68,

41

9 Bhatinda 440 0.699 0.7 IRS 0.998 41, 19, 68, 1

10 BhatindaExt. 420 0.972 0.973 IRS 0.999 68, 14, 19, 1, 41

11 Bhusawal 483 0.821 0.828 IRS 0.992 59, 46, 19

12 Birasinghpur 840 0.703 0.703 - 1 73, 68, 59, 19


12

DMU No

DMU Capacity CRS TE

VRS TE


Peers

13 Bokaro B 630 0.667 0.668 IRS 0.999 1, 41, 19, 68

14 Budge Budge 500 1 1 - 1 14

15 Calcutta 160 0.671 0.945 IRS 0.71 41, 52, 61

16 Chandrapur 2340 0.8 0.802 DRS 0.997 59, 53, 64, 68

17 Chandrapura 780 0.73 0.731 IRS 0.998 19, 41, 53, 68

18 Dadri 840 1 1 - 1 18

19 Dahanu 500 1 1 - 1 19

20 Durgapur 350 0.719 0.737 IRS 0.975 1, 41, 19, 68, 52

21 Durgapur

(DPL)

395 0.687 0.707 IRS 0.971 68, 41, 67, 19,

61

22 Ennore 450 0.524 0.53 IRS 0.989 1, 41, 19, 68

23 Farakka

STPS

1600 0.896 0.902 DRS 0.993 19, 59, 73

24 Faridabad 180 0.559 0.636 IRS 0.88 19, 52, 68, 41,

61

25 Gandhinagar 660 0.829 0.83 - 1 41, 19, 1, 68

26 Harduaganj 450 0.46 0.466 IRS 0.987 41, 1, 68

27 I.P.Stn. 248 0.52 0.571 IRS 0.911 1, 52, 41, 19, 68

28 IB Valley 420 0.86 0.868 IRS 0.991 46, 59, 19, 14

29 Kahalgaon 840 0.927 0.936 DRS 0.991 71, 73, 59, 14

30 Khaperkheda 840 0.796 0.801 DRS 0.994 14, 19, 73

31 Kolaghat 1260 0.737 0.789 DRS 0.935 53, 19, 68

32 Koradi 1100 0.721 0.731 DRS 0.987 19, 1, 68, 53

33 Korba East 440 0.839 0.844 IRS 0.994 46, 14, 19

34 Korba STPS 2100 1 1 - 1 34

35 KorbaWest 840 0.796 0.806 DRS 0.987 19, 14, 73

36 Kota 1045 0.916 0.973 DRS 0.942 53, 71, 73, 14

37 Kothagudem 1180 0.813 0.836 DRS 0.972 14, 73

38 Mejia 840 0.82 0.864 DRS 0.949 53, 73, 1

39 Mettur 840 0.939 0.942 DRS 0.997 59, 19, 73

40 Nasik 910 0.788 0.803 DRS 0.982 53, 19, 68, 1

41 Nellore 30 1 1 - 1 41

42 North

Chennai

630 0.827 0.827 - 1 68, 41, 1, 19

43 Obra 1550 0.589 0.595 DRS 0.991 19, 68, 1

44 Panipat 1360 0.743 0.781 DRS 0.952 53, 73, 68, 1

45 Panki 220 0.614 0.719 IRS 0.854 52, 41, 61, 68

46 Paras 63 1 1 - 1 46

47 Paricha 430 0.543 0.55 IRS 0.986 1, 41, 68, 19

48 Parli 690 0.858 0.871 DRS 0.985 73, 14

49 Patratu 840 0.505 0.509 IRS 0.992 41, 1, 68

50 Raichur 1470 0.826 0.869 DRS 0.95 19, 53, 68

51 Rajghat 135 0.562 0.652 IRS 0.861 19, 41, 1, 68, 52

52 Ramagundam 62 0.792 1 IRS 0.792 52

53 Ramagundam

STPS

2600 1 1 - 1 53

54 Rihand 2000 0.963 0.982 DRS 0.981 53, 73, 68, 19

55 Ropar 1260 0.86 0.906 DRS 0.949 19, 73, 1, 53

56 Santaldih 480 0.714 0.756 IRS 0.944 1, 41, 52, 68, 2

57 Satpura 1143 0.798 0.803 DRS 0.994 59, 19, 73

58 Sikka 240 0.772 0.776 IRS 0.995 19, 41, 1, 68


13

DMU No

DMU Capacity CRS TE

VRS TE


Peers

59 Simhadri 1000 1 1 - 1 59

60 Singrauli 2000 0.946 0.957 DRS 0.989 59, 73, 19

61 Southern

Repl

135 1 1 - 1 61

62 Suratgarh 1250 0.926 0.991 DRS 0.935 14, 53, 71, 73

63 Talcher 470 0.861 0.864 IRS 0.996 14, 19, 46

64 TALCHER-

Kaniha

3000 1 1 - 1 64

65 Tanda 440 0.867 0.872 IRS 0.995 14, 46

66 Tenughat 420 0.704 0.706 IRS 0.997 41, 19, 1

67 Titagarh 240 0.964 1 IRS 0.964 67

68 Trombay 1150 1 1 - 1 68

69 Tuticorin 1050 0.87 0.878 DRS 0.992 19 ,14, 73

70 Ukai 850 0.816 0.816 - 1 41, 68, 53, 19

71 Unchahar 840 0.981 1 DRS 0.981 71

72 Vijayawada 1260 0.898 0.989 DRS 0.907 73, 53, 71

73 Vindhyachal 2260 1 1 - 1 73

74 Wanakbori 1260 0.809 0.849 DRS 0.953 19, 53, 73, 1

Table 2: Technical Efficiency, Scale Efficiency and Return to Scale

2. While the efficiency frontier defined by the operational parameters consisting of

Generation, installed capacity, capacity unutilized because of PM and FO, coal and oil

consumption, without relaxing for scales of operation (CRS frontier) is occupied by 11

power plants, after relaxing for the unique scales of operation 4 more power plants move to

the VRS efficiency frontier. The list of efficient plants is detailed in table 3.

Sector Operator Plants

Reliance Energy Dahanu

Tata Power Trombay Private – 5 Plants

[Total - ] CESC Budge Budge, Titagarh

*, SouthGen.

Central – 7 Plants

[Total - ]

NTPC Vindhyachal, Unchahar*,

Ramagundam, Talcher – Kaniha,

Simhadri, Dadri, Korba

GSECL Ahmedabad

APGenco Ramagundam*, Nellore

MahaGenco Paras

State – 5 Plants

[Total - ]

MPGenco Amarkantak*

Table - 3: Power plants defining the efficiency frontier (*

occupy only the VRS frontier)

3. These plants have VRS technical efficiency of 100%. It can be seen that even

though four plants have CRS Technical Efficncy as low as 53.0% (Amarkantak) , 79.2%


14

(Ramagundam, 96.4% (Titagarh) and 98.1% (Unchahar), these plants occupy the efficiency

frontier because there are no other plants of comparable size with whom their performance

could be compared. It is found that 450MW Harduaganj plant has the lowest efficiency of

46.6% followed by 320MW Barauni 47.9% and the efficiency of only these two plants are

below 50%. The efficiency of other remaining plants varies in the range of 50.9% to 100%

with mean value of 83.9%. Dahanu, which is an efficient plant, has the highest peer count of 38

i.e. 38 other power stations see best practices in Dahanu compared to 32 of Trombay. The peer

count of other efficient DMUs is shown in figure-2.

0

5

10

15

20

25

30

35

40

Ah

em

ad

ab

ad

Am

ark

an

tak

Bu

dg

eb

ud

ge

Da

ha

nu

Ne

llore

Pa

ras

Ra

ma

gu

nd

am

Ra

ma

gu

nd

am

ST

PS

Sim

ha

dri

So

uth

ern

Re

pl

Ta

lch

er

-

Ka

nih

a

Tita

ga

rh

Tro

mb

ay

Un

cha

ha

r

Vin

dh

yach

al

Figure-2: Peer Count of Efficient DMUs

4. It is observed that 18 plants have constant return to scale, 27 plants have

decreasing return to scale and 29 plants have increasing return to scale.

5. Operator wise analysis of the efficiency scores reveal that while Reliance Energy

and Tata Power have an average efficiency of 100%, the average efficiency level of 4 other

operators – CESC, RRVUNL , APGenCo and NTPC is above 95%. The average efficiency

of BSEB is lowest at 47.9% followed by JSEB at 50.9%. The average efficiency of as

many as 13 operators is below the national average. Operator wise average efficiency is

detailed in figure-3.


15

Average Efficiency Scores of Different Operators

0.450

0.550

0.650

0.750

0.850

0.950

AP

GE

NC

O

BS

EB

CE

SC

CS

EB

DP

L

DV

C

GS

EC

L

HP

GC

L

IPG

PC

L

JS

EB

KP

CL

MA

HA

GE

NC

O

MP

GP

CL

NT

PC

OP

GC

PS

EB

RE

L

RR

VU

NL

TA

TA

PC

L

TN

EB

TV

NL

UP

RV

UN

L

WB

PD

C

Operator

VR

S T

E

Figure-3: Average Efficiency Score of different Operators

6. The variation of average efficiency based on plant size is shown in figure-4.

Average Efficiency

0.750

0.800

0.850

0.900

0.950

1.000

1.050

30 to 500 500 to 1000 1000 to 1500 1500 to 2000 2000 to 2500 2500 to 3000 3000 or More

Figure-4: Variation of Average Efficiency with Plant Capacity

It can be seen that barring plant capacities in the 1500 to 2000 MW band, the average

efficiency increases with plant size. There are 3 plants in the 1500MW to 2000MW band

namely Obra – 1550MW, Farakka STPS – 1600 MW and Anpara – 1630 MW and the

efficiency scores are 59.5%, 90.2% and 86.4% respectively. The average efficiency of

plants of size 2500MW or higher is 100%. As such there are 2 plants Ramagundam STPS –

2600MW and Talcher Kaniha – 3000MW and both lie on the CRS efficient frontier. The

smallest and oldest plant the 30 MW Nellore plant has peer count of 25 and turns out to be

efficient in both the CRS as well as VRS frontiers. This is inline with the findings of

Golany (1994) and Diewert and Nakamura (1999).

7. The variation of average efficiency with average unit size is shown in figure-5.


16

0.600

0.650

0.700

0.750

0.800

0.850

0.900

0.950

1.000

1.050

30 to 100 MW 50 to 149 MW 150 to 249MW 250 to 349MW 350 to 449MW 450MW and above

Figure-5: Variation of Average Efficiency with Average Unit Size

It is seen that barring plants with average unit size in the 50-149 MW band, the CRS

efficiency increases with unit size.

8. Sector wise performance analysis of the power plants reveal, average efficiency of

the plants in Private Sector has the highest average efficiency of 99.08% followed by plants

in Central Sector and State Sector at 91.03% and 79.48% respectively. The findings are in

line with the observations “the private plants have higher technical and scale efficiencies

which hint at better managerial skills of the private sector” - Sarica (2006), and “the

average efficiencies for the private plants are higher for the two developing country

groupings (Caribbean and Tanzania)...” - Diewert and Nakamura (1999).

9. Region wise analysis of average efficiency scores indicate that the plants in

Southern Region have the highest average efficiency of 89.73% followed by Western

Region at 85.88% and Eastern Region at 82.12%. Plants in the Northern Region have the

lowest average efficiency of 80.30%. The state wise plot of average efficiency is shown in

figure-6. It is seen that the plants in Rajasthan highest average efficiency score of 98.2%

followed by those in Andhra Pradesh at 97.1% and Orissa at 91.1%. The plants in

Jharkhand, Delhi and Bihar are least efficient.


17

State wise Plot of Average Efficiency

0.6

0.7

0.8

0.9

1RAJASTHAN

ANDHRA PRADESH

ORISSA

CHATTISGARH

WEST BENGAL

MAHARASHTRA

KARNATAKA

PUNJAB

GUJARAT

MADHYA PRADESH

UTTAR PRADESH

TAMIL NADU

HARYANA

BIHAR

DELHI

JHARKHAND

Figure-6: State wise variation of average Efficiency

10. Efficiency plot of different plants is shown in figure-7 and summary distribution of generation capacity and number of

plants in different efficiency bands is shown in table-4.

Efficiency Plot of Different Plants

0.5

0.6

0.7

0.8

0.9

1Ahemadabad

AmarkantakAmarkantakExtAnparaBadarpur

BandelBarauni

BarkeshwarBhatinda

BhatindaExt.

Bhusawal

Birsingpur

Bokaro B

BudgeBudge

Calcutta

Chandrapur

Chandrapura

Dadri

Dahanu

Durgapur

Durgapur(DPL)

Ennore

Farakka STPS

Faridabad

Gandhinagar

Harduaganj

I.P.Stn.

IbValley

Kahalgaon

KhaperkhedaKolaghat

KoradiKorbaEast

Korba STPSKorbaWestKotaKothagudem

MejiaMetturNasikNellore

NorthMadrasObra

PanipatPanki

Paras

Paricha

Parli

Patratu

Raichur

Rajghat

Ramagundam

Ramagundam STPS

Rihand

Ropar

Santaldih

Satpura

Sikka

Simhadri

Singrauli

Southern Repl

Suratgarh

Talcher

Talcher -Kaniha

Tanda

Tenughat

TitagarhTrombay

TuticorinUkai

UnchaharVijayawadaVindhyachalWanakbori

Figure-7: Efficiency plot of Different Plants

Plants Generation Capacity Efficiency Band

(%) Number % age MW % age

Below 60 7 9.46 4288 6.88

60 – 70 4 5.41 1185 1.90

70 – 80 12 16.22 7885 12.65

80 – 90 22 29.73 19846 31.85

90 – 95 5 6.76 4700 7.54

Above 95 24 32.43 24405 39.17

Table-4: Distribution of Plants and Installed Capacity in Different Efficiency Bands

6 Conclusions:

We have attempted to model the relative performance level of coal fired thermal power

plants in India during 2005-06 based on as many as 6 inputs and one output. The TE of

plants varies from 46.6% to 100%. It is seen that the mean TE with CRS auumptions is

80.9% and with VRS is 83.9%. Out of the 74 power plants, the technical efficiency

(VRS) of as many as 34 plants having aggregate capacity of 23,774MW is below the

mean TE of 83.9%. This indicates substantial scope for contraction of the current input

levels without deteriorating the output levels. Lesser consumption of inputs will not only

reduce the cost of electricity generation there by enhancing the competitiveness but also

make available the scarce inputs to generate more and more electricity. State wise

analysis of the average TE indicates power plants in Rajasthan are most efficient

followed by Andhra Pradesh and Orissa. The plants in Jharkhand, Delhi and Bihar are

least efficient. Plants in the Southern Region are most efficient followed by those in

Western and Eastern Region. The plants in the Northern Region are least efficient.

Considering the size and importance of the sector, it warrants more detailed productivity

studies like analyzing the productivity trend over a 5-10 years horizon, extension of the

study to unit level by capturing more and more parameters and validation of the findings

with the field professionals. More and more real life constraints could be incorporated to

model as close as to the real world business environment.


20

References:

1. Banker R D, Cooper W and Charnes A (1984), Some models for estimating

technical and scale inefficiency in data envelopment analysis, Management Science 30,

pp 1078-1092.

2. Cooper W , Seiford L.M. and Tone K(2007), Data Envelopment Analysis,

Springer Science + Business Media

3. Ramanathan R (2003), An Introduction to Data Envelopment Analysis – A Tool

for Performance Management, Sage Publications India Pvt. Ltd, New Delhi.

4. Golany Boaz, Yaakov Roll, and David Rybak (1994), Measuring Efficiency of

Power Plants in Israel bv Data Envelopment Analysis, IEEE Transactions On

Engineering Management, VOL. 41, NO. 3, AUGUST 1994, pp 291-301.

5. Emrouznejad, Ali (2001), An extensive bibliography of data envelopment

analysis (DEA), Volume I: Working Papers, Business School, University of Warwick,

Coventry CV47AL, England.

6. Olatubi W.O and Dismukes D.E (2000), A data envelopment analysis of the

levels and determinants of coal-fired electric power generation performance, Utilities

Policy 9 , pp 47-59.

7. Sueyoshi T and Goto M (2001), Slack Adjusted DEA for time series analysis:

Performance measurement of Japanese electric power generation industry in 1984-1993,

European Journal of Operational Research 133, pp 232-259.

8. Lam Pun-Lee and Shiu Alice (2001), A data envelopment analysis of the

efficiency of China’s thermal power generation, Utilities Policy 10, 75-83.

9. Shanmugan K.R. and Kulshreshtha P (2005), Efficiency analysis of coal-based

thermal power generation in India during post-reform era , Intl J Global Energy Issues,

Vol. 23, No 1, pp 15-28.

10. Cook W.D and Green R.H (2005), Evaluating power plant efficiency: a

hierarchical model, Computers & Operations Research, 32 , pp-813-823.

11. Vaninsky A (2006), Efficiency of electric power generation in the United States:

Analysis and forecast based on data envelopment analysis, Energy Economics 28, pp

326-338.


21

12. Dash A.P, Behera S.K. and Rath B.P. (2008), Alternative Matrices for Powering

India’s Power demands, Atoms For Peace : An International Journal, Vol 2, No 1, 2008

13. Website of Central Electricity Authority (CEA) http://www.cea.nic.in - Last

visited on 15-05-2009

14. Website of Ministry of Power (MoP) http://www.powermin.nic.in/ - Last visited

on 12-05-2009.

15. Coelli, T. A Guide to DEAP Version 2.1: A Data Envelopment Analysis

(Computer) Program, CEPA Working Paper 96/08, Centre for Efficiency Analysis and

Productivity Analysis, University of New England, Australia.

16. Yang Hongliang and Pollitt Michael (2007), Incorporating Both Undesirable

Outputs and Uncontrollable Variables into DEA: the Performance of Chinese Coal-Fired

Power Plants,

17. Andersen P and Petersen N.C (1993), A procedure for ranking efficient units in

DEA, Management Science 39, pp. 1261-1264,1993.

18. Toba Natsuko (2003) , ERD Working Paper No 44, Welfare Impacts of Electricity

Generation sector Reform in the Philippines, Asian Development Bank.

19. Hirschhausen, CV, Cullman A and Kappeler A (2006), Efficiency analysis of

German electricity distribution utilities – non-parametric and parametric tests, Applied

Economics, 38 pp 2553-2566.

20. Sarica Kemel and Or Ilhan (2006), Efficiency Assessment of Turkish power

plants using data envelopment analysis, Energy.

21. Sumanth, D.J.(1998) , Total Productivity Management : a Systematic and

Quantitative Approach to Compete in Quality, Price and Time, Boca Raton, Florida : St.

Lucie Press

22. Kumbhakar, S.C and Lovell C.A.K (), Stochastic Frontier Analysis, Cambridge

University Press

23. Kopp R.J and Smith V.K.(1980), Frontier Production Estimates for Steam Electric

Generation : A Comparative Analysis, Southern Economic Journal, Vol. 46, No. 4 (Apr

1980), pp 1049-1059.


22

24. Annual Thermal Performance Review 2005-06 (2006), Central Electricity

Authority, New Delhi. Also available at

http://cea.nic.in/god/opm/Thermal_Performance_Review/0506/COVER%20PAGE.pdf .

25. Thanassoulis E, Boussofiane A and Dyson R.G.(1996), A Comparison of Data

Envelopment Analysis and Ratio Analysis as Tools for Performance Assesment, Omega

Int. Journal of Mgmt Science, Vol 24, No 3, pp. 229-244.

26. Proposed comprehensive award scheme for power sector (2008), Central

Electricity Authority, New Delhi available at

http://www.cea.nic.in/Thermal/project_monitoring/Special%20Reports_TPM/Proposed_r

evised_Scheme-07_08.pdf - last visited on 20-05-2009.

27. Malhotra D.K. and Malhotra R. (2008), Analyzing Financial Statements using

Data Envelopment Analysis,

28. Wu Caicong, Chen Xiuwan and Yang Yinsheng (2004), Decision-making

modeling method based on artificial neural network and data envelopment analysis,

Geoscience and Remote Sensing Symposium, IGARSS apos; 04.Proceedings 2004 IEEE

International Volume 4, Issue, 20-24 Sept 2004 Pages 2435-2438.

29. Emrouznejad Ali and Shale Estelle (2009) , A combined neural network and

DEA for measuring efficiency of large scale datasets, Computers and Industrial

engineering, Vol 56, Issue 1, Page- 249.

30. Celebi Dilay and Bayraktar Demet (2008), An integrated neural network and data

envelopment analysis for supplier evaluation under incomplete information, Expert

Systems with Applications: An International Journal, Vol 35, Issue 4, Pages 1698-1710.

31. Adler, N and Golany B. (2001), Evaluation of deregulated airline networks using

data envelopment analysis combined with principal component analysis with an

application to Western Europe, European Journal of Operational Research , 132 (2)

,Pages 18-31

32. Adler N and Yazhemsky E (2009), Improving Discrimination in Data

envelopment Analysis: PCA-DEA or Variable Reduction, European Journal of

Operational Research, In Press, Corrected Proof, Available online 7 April 2009


23

33. Kuosmanen Timo (2006), Stochastic Nonparametric Envelopment of Data:

Combining virtues of SFA and DEA in a Unified Framework, MTT Discussion Papers 3,

Economic Research Unit, MTT Agrifood Research Finland.

34. Kao Chiang and Liu Shiang-Tai (2000), Fuzzy efficiency measures in data

envelopment Analysis, Fuzzy Sets and Systems, 113 , pp 427-437

35. Guo Peijun and Tanaka Hideo (2001), Fuzzy DEA: A perceptual evaluation

method, Fuzzy Sets and Systems, 119(2001), pp 149-160

36. Li, S X.(1998), Stochastic models and variable returns to scales in data

envelopment analysis, European Journal of Operational Research, 104, 532-548

37. Huang, Z., and S.X. Li (2001), Stochastic DEA Models With Different Types of

Input-Output Disturbances, Journal of Productivity Analysis, 15, 95-113

38. Mishra, S. K., A Brief History of Production Functions (October 9, 2007).

Available at SSRN: http://ssrn.com/abstract=1020577

39. Shephard, R.W. (1970), Theory of Cost and Production Functions, Princeton

University Press, Princeton, NJ.

Documents

Performance Evaluation and Efficiency Analysis of …banker/dea2009/paper/Behera.pdfIndian economy, which till recently ... Performance evaluation and efficiency analysis of the thermal